Hospital Outcomes of Infants with Neonatal Opioid Withdrawal Syndrome at a Tertiary Care Hospital with High Rates of Concurrent Nonopioid (Polysubstance) Exposure

Abstract

Objective

Neonatal opioid withdrawal syndrome (NOWS) describes infants’ withdrawal signs and symptoms after birth due to an interruption of prenatal opioid exposure. Many infants with NOWS are also exposed to nonopioids, however. This study was to determine hospital outcomes of infants exposed to opioids alone or coexposed with nonopioid substances (polysubstance).

Study Design

We reviewed infants of ≥34 weeks of gestation with prenatal opioid exposure from April 2015 to May 2018. We compared the median lengths of stay (LOS) and treatment (LOT) and the percentages of infants requiring pharmacologic and adjunctive treatment in infants exposed to opioids alone or polysubstance. We used Wilcoxon’s test for continuous outcomes or Chi-squared test for categorical outcomes to determine statistical significance. We used multivariable regression model to calculate each drug category’s estimates of adjusted mean ratios for LOS and LOT plus estimates of adjusted odds ratios for pharmacologic/adjunctive treatments.

Results

Of the 175 infants, 33 (19%) infants had opioid exposure alone. Opioid exposure included short- and/or long-acting opioids. A total of 142 (81%) had polysubstance exposure with 47% of mothers using nicotine products. We saw similar hospital outcomes between infants exposed to opioids alone or polysubstance; however, a higher percentage of infants with both short- and long-acting opioid exposure required pharmacologic treatment compared with either opioid alone. Focusing on individual drug categories, we detected differential hospital outcomes in which short-acting opioids decreased LOT, whereas long-acting opioids increased LOS, LOT, and need for pharmacologic and adjunctive treatment. Coexposure of opioids with stimulants decreased LOT and reduced need for adjunctive treatment. Coexposures with antidepressants increased LOT, while with antiepilepetics increased LOS.

Conclusion

Because infants with NOWS often have coexposures to other nonopioid substances, appreciating the associated risks of individual or combination of drugs in modulating hospital outcomes may help counsel families on their infants’ expected hospital course.

In the United States, the rates of opioid use disorder (OUD) at delivery hospitalization have more than quadrupled during 1999 to 2014.¹ As a result, we are witnessing an increasing rate of infants who experience neonatal opioid withdrawal syndrome (NOWS). NOWS is a constellation of gastrointestinal, autonomic, and central nervous system signs and symptoms after birth due to an interruption of prenatal opioid exposure.² Even though NOWS classically refers to opioid withdrawal, many mothers with OUD use other illicit, licit, and prescribed substances that could exacerbate the signs and symptoms of NOWS, potentially affecting their infants’ hospital outcomes. For infants with mild withdrawal, we advocate nonpharmacologic maneuvers, such as minimizing stimuli and swaddling, as ways to improve the infants’ ability to eat, sleep, and gain weight. For infants with moderate-to-severe withdrawal, we add an opioid as pharmacologic treatment. We add adjunctive medications in addition to an opioid if withdrawal is not well controlled.

Determinants of the incidence and severity of withdrawal in infants with NOWS are not well understood. Specific genetic polymorphisms, pharmacological properties of exposures, and polysubstance exposure have all been described.^3,4 The effects of pure prenatal opioid exposure are the most studied; however, many infants are exposed to substances in addition to opioids as previously mentioned. Some studies have reported rates of polysubstance exposure in pregnancy as high as 65%,⁵ but we know little about how individual or combination of substances may affect NOWS outcomes in these infants. In this study, we set out to determine the effects of opioids alone (whether short- or long-acting or both) or coexposure of opioids with other substances (polysubstance) on length of stay (LOS), length and requirement for pharmacologic treatment, and requirement for adjunctive treatment in infants with NOWS. We additionally calculated the estimates of adjusted means ratios (AMRs) for LOS and length of treatment (LOT), as well as the adjusted odds ratios (AORs) for the requirement for pharmacologic including adjunctive treatment, based on the drug categories after adjusting for confounding variables.

Materials and Methods

Study Setting

This study was conducted at the University of Utah Hospital (UUH), a tertiary urban referral and academic medical center with approximately 5,000 deliveries per year. Infants with NOWS were treated in either our newborn/intermediate care nursery staffed by general academic pediatricians or our level-III NICU staffed by neonatologists. Infants were treated in the NICU only when other medical issues required ICU care or if they needed clonidine as an adjunctive due to need for more intensive monitoring of heart rate and blood pressure. The ability for parents to room-in was not available during this study period.

Care Process Model

We implemented a revised NOWS care process model in April 2015 to unify practices between the two units. It recommends observation for withdrawal based on maternal substance use history and positive toxicology screens from mother’s urine and infant’s urine, meconium, or umbilical cord screens. We use the Neonatal Withdrawal Inventory (NWI; Fig. 1) to score for signs/symptoms rather than the widely used modified Finnegan’s scoring tool because NWI has been demonstrated to accurately reflect neonatal withdrawal severity and has improved interrater reliability compared with the modified Finnegan.⁶ The NWI helps to differentiate infants with mild withdrawal symptoms who respond to nonpharmacologic treatment from those with moderate-to-severe withdrawal who require pharmacologic treatment. Scoring is done in every 3 to 4 hours during nursing care such as diaper change and feeding to reinforce nonpharmacologic interventions.

Fig. 1.

Neonatal withdrawal inventory (NWI) scoring tool.⁶ Scoring begins with a prestimulus observation period of 1-minute in which respiratory rate is recorded and whether tremors when undisturbed are present. The stimulus observation period assesses for tone, Moro, tremors when disturbed, sweating or mottling, vomiting/regurgitation, loose watery stools, temperature, sneezing or yawning while providing cares, feeding, and comfort measures. Once infant is reswaddled and positioned back into a crib, an additional 1-minute poststimulation observation period assesses for persistent irritability and crying. The maximum number of points attainable is 19.

For infants whose scores are below the treatment threshold (≤7), similar to the modified Finnegan, it recommends the use of nonpharmacologic interventions (such as swaddling with soft blankets, being in a dark quiet environment, clustering care to minimize stimulation, kangaroo care if parent is available, and use of a swing) for at least 72 hours for any short-acting opioid exposure or 96 hours for any long-acting opioid exposure. For infants with moderate-to-severe withdrawal identified as two consecutive scores ≥8, a 13-step weight-based morphine protocol is implemented for pharmacologic treatment. Weaning begins when two consecutive scores become ≤7. For infants with scores persistently ≥8 and are unable to wean, adjunctive treatment with either clonidine or phenobarbital is recommended. Clonidine is recommended for primarily opioid exposure whereas phenobarbital is recommended for polysubstance exposure with multiple drug classes. If morphine treatment is initiated and infant weans successfully according to the weaning schedule, the predicted LOT would be 8 days. Recommendation for earliest discharge is 24 hours after the last morphine dose. If clonidine was required for adjunctive therapy, this was weaned off prior to hospital discharge. If phenobarbital was required for adjunctive therapy, infants were discharged home after discontinuing morphine and phenobarbital was weaned by their outpatient pediatrician.

Data Collection

The study was approved by the UUH Institutional Review Board. A retrospective chart review of the electronic health record was conducted on all infants ≥34 weeks of gestation who had any prenatal opioid exposure between April 2015 to May 2018 and received NWI scoring. Maternal and infant exposures were extracted from history and physicals, social histories, discharge summaries, and toxicology results. The research team resolved conflicting exposures by verifying whether mother received the drug in question during labor or delivery; if not, the drug of question was included. We categorized the prenatal substance exposures as follows: short-acting opioids, long-acting opioids, tobacco/nicotine, stimulants, antidepressants, cannabinoids/phencyclidine (PCP), sedatives, antiepileptics, and antipsychotics. Short-acting opioids included morphine, heroin, fentanyl, oxycodone; long-acting opioids included methadone and buprenorphine; stimulants included methamphetamine, cocaine, and methylphenidate; antidepressants included selective serotonin reuptake inhibitors, serotoninnorepinephrine reupake inhibitors, bupropion, and duloxetine; cannabinoids referred to marijuana; sedatives included benzodiazepines, zolpidem, and trazodone; antiepileptics included levetiracetam, phenobarbital, gabapentin; and antipsychotics included olanzapine, quetiapine, and risperidone. We defined polysubstance exposure as exposure to opioids plus ≥1 nonopioid substances. We counted an infant with multiple exposures into each drug category.

We determined the following in-hospital NOWS outcomes: (1) LOS which included all infants receiving nonpharmacologic and pharmacologic treatment, (2) requirement for pharmacologic treatment, (3) LOT defined as when a medication was started to when a medication was discontinued, and (4) requirement for adjunctive medication. LOT referred to the number of days on morphine and clonidine in infants who required clonidine as an adjunct, whereas it referred to the number of days on morphine only for infants who required phenobarbital as an adjunct.

Statistical Analyses

We consulted with the University of Utah, Center for Clinical and Translational Science Study Design and Biostatistics Center, for our statistical analyses. We compared the effects of opioids or polysubstance on NOWS outcomes using the Wilcoxon’s test for continuous outcomes or the Chi-squared test for categorical outcomes. We used a multivariable regression model to calculate the estimates of AMRs for LOS and LOT and the estimates of AORs for pharmacologic and adjunctive treatments of each exposure. Due to the nonnormal distribution for LOS and LOT, we used a generalized linear model with a gamma distribution and log-link function. For pharmacologic and adjunctive treatment outcomes, logistic regression was utilized. Each model was controlled for confounding variables that could affect outcomes including breastfeeding at discharge (particularly important if mother was on maintenance assisted therapy with long-acting opioids), gestational age, small for gestational age (defined as birth weight ≤10% for gestational age), and any other medical or social complications (such as need for prolonged antibiotics or Child Protective Services involvement) that could influence outcomes. We declared statistical significance with p ≤ 0.05 for all analyses.

Results

Hospital NOWS Outcomes Comparing Exposure to Opioids Alone or Polysubstance

A total of 175 infants met our inclusion criteria. Table 1 shows the descriptive statistics of our study cohort. Important to highlight is that 19% was exposed to opioids alone while 81% was exposed to polysubstance that included one opioid plus at least one nonopioid. Because one infant could have had multiple drug exposures, the total number of infants identified under “Drug categories” exceeded 175 infants. The median LOS for opioid or polysubstance exposure was similar at 7 (Q1–3: 4–14) and 10 (Q1–3: 4–14) days, respectively (Table 2). We additionally noted similar median LOT (10 days each) and need for pharmacologic treatment (46% each) including adjunctive (15–16%) between the two groups (Table 2).

Table 1.

Descriptive statistics of our study cohort of ≥34 weeks of gestation with NOWS (n = 175, includes prenatal opioid exposure alone or polysubstance)

Demographics	n (%) or median (Q1–Q3)
Male sex	96 (55)
Gestational age at delivery (wk)	38.7 (37.3–39.3)
Inborn	161 (92)
Discharged from newborn or intermediate care Nursery/NICU	140 (80)/35 (20)
Breastfeeding at discharge	84 (48)
Small for gestational age (birth weight ≤10% for age)	42 (24)
Prenatal exposures	n (%)
Polysubstance (opioids + ≥ 1 nonopioid) Including Opioids	142 (81)
Opioids alone	33 (19)
Drug categoriesa	n (%)
Short-acting opioids	124 (71)
Long-acting opioids	107 (61)
Tobacco/nicotine (any formulation)	83 (47)
Stimulants (cocaine and amphetamine-type substances)	66 (38)
Antidepressants	39 (22)
Cannabinoids/phencyclidine (PCP)	33 (19)
Sedatives	31 (18)
Antiepileptics	14 (8)
Antipsychotics	10 (6)

Abbreviation: NOWS, neonatal opioid withdrawal syndrome.

^aAn infant may be counted in multiple drug categories therefore total number of infants in all categories exceed 175.

Table 2.

Hospital NOWS outcomes of all opioid-exposed infants (n = 175)

Hospital NOWS Outcomes	Opioids alone (n = 33) Median (Q1–Q3) or n (%)	Polysubstance (n = 142) Median (Q1–Q3) or n (%)	p-Value
Length of stay (days, all infants receiving nonpharmacologic and pharmacologic treatment)	7 (4–14)	10 (4–14)	0.88
Infants requiring pharmacologic treatment	15 (46)	65 (46)	0.97
Length of treatment (days, all infants requiring pharmacologic treatment)	10 (8–12)	10 (8–15)	0.79
Infants requiring adjunctive treatment (clonidine or phenobarbital)	5 (15)	22 (16)	0.96

Hospital NOWS Outcomes Comparing Types of Opioid Exposure

Of the 175 infants in our cohort, 68 had short-acting opioid exposure, 51 had long-acting opioid exposure, and 56 had both short- and long-acting opioid exposures. The median LOS and LOT were similar across these opioid subtypes (Table 3). However, infants with both short- and long-acting opioid exposures had increased requirement for pharmacologic treatment compared with either opioid individually (Table 3). The need for adjunctive treatment was similar across all opioid types.

Table 3.

Descriptive statistics of NOWS outcomes comparing types of opioids (n = 175)

Hospital NOWS outcomes	Short-acting opioids Median (Q1–Q3) or n (%)	Long-acting Opioids Median (Q1–Q3) or n (%)	Both short- and Long-acting opioids Median (Q1–Q3) or n (%)	p-Value
Length of stay (d)	8 (3–13) (n = 68)	5 (4–16) (n = 51)	12 (5–15) (n = 56)	0.110
Pharmacologic treatment	25 (37)	20 (39)	35 (63)	0.009
Length of treatment (d)	9 (8–15) (n = 25)	11.5 (9–21) (n = 20)	10 (8–12) (n = 35)	0.051
Need for adjunctive treatment	6 (8.8)	11 (22)	10 (18)	0.135

Abbreviation: NOWS, neonatal opioid withdrawal syndrome.

Multivariable Logistic Regression Model Comparing Drug Categories on NOWS Outcomes

We performed a multivariable logistic regression model to examine the associations of individual drug categories on hospital NOWS outcomes after adjusting for confounding factors. Exposure to short-acting opioids was associated with a 31% decrease in LOT (AMR = 0.69 [0.55–0.87], Table 4). Conversely, exposure to long-acting opioids was associated with a 53% increase in LOS (AMR = 1.53 [1.22–1.91], Table 4) and a 29% increase in LOT (AMR = 1.29 [1.02–1.63], Table 4). Exposure to long-acting opioids was also observed to have a 2.5-fold increased need for pharmacologic treatment (AOR = 2.53 [1.27–5.06], Table 4) and 3.2-fold increased need for adjunctive treatment (AOR = 3.23 [1.16–−9.01], Table 4). Coexposure of opioids with stimulants showed a 23% decrease in LOT (AMR = 0.77 [0.63–−0.95], Table 4) plus a 78% reduction in the need for adjunctive treatment (AOR = 0.22 [0.08–0.67], Table 4). Coexposure of opioids with antidepressants showed a 38% increase in LOT (AMR = 1.38 [1.09–1.74], Table 4). Lastly, coexposure of opioids with antiepileptics had a 50% increase in LOS (AMR = 1.50 [1.02–2.21], Table 4). We found no associations in NOWS outcomes for concurrent exposures to tobacco/nicotine, cannabinoids/PCP, sedatives, or antipsychotics with opioids.

Table 4.

Multivariable logistic regression model examining the associations of individual drug categories on hospital NOWS outcomes (n = 175) after adjusting for confounding variables^a

Drug categories	Length of stay		Pharmacologic treatment		Length of treatment		Adjunctive treatment
	Estimates of adjusted means ratio Lcl-Ucl	p-Value	Estimates of adjusted odds ratio Lcl-Ucl	p-Value	Estimates of adjusted means ratio Lcl-Ucl	p-Value	Estimates of adjusted odds ratio Lcl-Ucl	p-Value
Short-acting opioids	0.87 (0.69–1.10)	0.245	1.15 (0.57–2.33)	0.688	0.69 (0.55–0.87)	0.002	0.41 (0.16–1.03)	0.059
Long-acting opioids	1.53 (1.22–1.91)	0.0002	2.53 (1.27–5.06)	0.009	1.29 (1.02–1.63)	0.034	3.23 (1.16–9.01)	0.025
Tobacco/nicotine	0.89 (0.72–1.10)	0.269	0.61 (0.32–1.16)	0.132	0.87 (0.70–1.08)	0.203	0.71 (0.30–1.68)	0.435
Stimulants	0.81 (0.65–1.00)	0.051	1.00 (0.52–1.94)	0.997	0.77 (0.62–0.95)	0.013	0.22 (0.08–0.67)	0.007
Antidepressants	1.25 (0.97–1.61)	0.082	1.33 (0.64–2.77)	0.453	1.38 (1.09–1.74)	0.008	2.30 (0.93–5.65)	0.070
Cannabinoids/PCP	0.93 (0.71–1.22)	0.622	1.20 (0.55–2.61)	0.650	0.83 (0.64–1.09)	0.177	0.99 (0.34–2.90)	0.991
Sedatives	1.01 (0.77–1.32)	0.965	0.77 (0.34–1.73)	0.523	1.25 (0.94–1.66)	0.121	0.79 (0.25–2.51)	0.685
Antiepileptics	1.50 (1.02–2.21)	0.042	2.28 (0.70–7.45)	0.172	1.14 (0.81–1.60)	0.454	1.82 (0.45–7.44)	0.402
Antipsychotics	1.38 (0.87–2.16)	0.167	1.87 (0.49–7.20)	0.363	0.96 (0.64–1.46)	0.864	1.42 (0.27–7.40)	0.676

Abbreviations: Lcl, lower 95% confidence limit; Ucl, upper 95% confidence limit; NOWS, neonatal opioid withdrawal syndrome; PCP, phencyclidine.

^aControlled for breastfeeding at discharge, gestational age, small for gestational age, presence of extraneous circumstances that delayed discharge such as need for prolonged antibiotics or Child Protective Services involvement.

Discussion

Using our new care process model in a single urban academic center, we discovered several important findings about the hospital outcomes of our infants with NOWS with either prenatal exposure to opioids alone or polysubstance. First, we found a disproportionately high rate of polysubstance exposure in our cohort. Despite this, opioid exposure alone had the same NOWS outcomes as polysubstance exposure, alluding to opioids being the main driver of hospital outcomes. In other words, nonopioid exposure in polysubstance did not worsen hospital NOWS outcomes from opioids alone. We also noted that the combination of short- and long-acting opioids significantly increased the need for pharmacologic treatment compared with either opioid alone. When individual drug combination was examined, we found that short-acting opioids, long-acting opioids, and coexposure of opioids with stimulants, antidepressants, and antiepileptics had various independent associated adjusted risks for hospital NOWS outcomes.

Since the surge of the opioid epidemic in the early 2000s, pregnant women represent a group whose opioid use has continued to increase. According to the 2017 National Survey on Drug Use and Health report, the use of opioids in the past month among pregnant women aged 15 to 44 years has doubled from 0.8% in 2015 to 1.6% in 2017.⁷ Interestingly, Utah ranks third in the United States for opioid prescription deaths⁸ and Utah pregnant women on Medicaid have a disproportionately high rate of prescription opioid dispensing during pregnancy at 42% compared with the national average of 22%.⁹ It is in this context that NOWS has become a commonplace in our nurseries. However, our population of NOWS infants are not only exposed to opioids; we have a high rate of polysubstance exposure (81%) at UUH with some of our mothers being exposed to ≥3 neuromodulating substances during pregnancy.¹⁰

Many studies have emerged in the last decade describing hospital outcomes of NOWS infants across the country.^11–17 Our finding are similar to these studies in that prenatal opioid exposure prolongs an infant’s typical length of hospital stay. Almost half of the NOWS infants in our study required pharmacologic treatment with morphine to temper their withdrawal while a smaller proportion required adjunctive treatment with either clonidine or phenobarbital. Splitting further into the types of opioids, the combined exposure of short- and long-acting opioids significantly increased the need for pharmacologic treatment, which undoubtedly would increase hospital costs despite no change in LOS or LOT. In the state of Utah, maintenance-assisted therapy with long-acting opioids in pregnant women with OUD has lagged, therefore we speculate that the high rate of short-acting opioid use is the result of maternal relapse to temper their withdrawal. Ideally, an interdisciplinary approach to the treatment of OUD during pregnancy is the best done by aligning obstetricians and substance use treatment specialists with ancillary support services. Such a comprehensive treatment program has been rare but is certainly improving and has become a priority in the combat of OUD in our state. We certainly advocate for the use of long-acting opioids to prevent relapse in mothers despite the associated increases in all NOWS outcomes examined.

Based on our clinical experience, we anticipated that prenatal use of nonopioid substances in addition to opioids would negatively affect NOWS outcomes. Indeed, we found that coexposure of opioids with antiepileptics or antidepressants was associated with increases in LOS or LOT. This is in line with a study by Wachman et al in which they found that mothers with OUD who are on concurrent psychiatric medications were associated with worse hospital NOWS outcomes.¹⁸ Another study by Huybrechts et al also showed significant increases in absolute risks of NOWS after intrauterine exposure to both opioids and antidepressants, antipsychotics, benzodiazepines, and gabapentin.¹⁹ Even though we found no association of cocurrent opioid and benzodiazepine exposure, a recent retrospective cohort study by Sanlorenzo et al reported that benzodiazepine exposure was an independent predictor of an infant developing pharmacologically treated NOWS.²⁰ We were surprised to find that coexposure of opioids with stimulants decreased LOT and the need for adjunctive treatment. The underlying basis for this is unclear but we noted that more than half of the maternal stimulant use accompanied short-acting opioid rather than long-acting opioid. We have additionally shown here that short-acting opioid decreased LOT, perhaps this observation in stimulants reflects the property of short-acting opioid rather than the stimulant itself. To our knowledge, this is the first description of the effects of combined opioid and stimulant use on NOWS hospital outcomes.

Limitations

We recognize certain limitations to this study. First, being retrospective in nature, the accuracy of data is dependent on the accuracy of the data entry and retrieval. To overcome this limitation, we examined multiple sources from the maternal and infant records to extrapolate prenatal exposures. The research team resolved conflicting exposures by verifying whether mother received the drug in question during labor or delivery; if not, the drug of question was included. We also reviewed all infants’ scores and their need for pharmacologic treatment individually to parse out adherence to the care process model. Second, because the majority of infants were exposed to multiple substances, our sample sizes in the opioids alone group plus in each individual exposure were small, potentially creating statistical errors that detected no differences in NOWS outcomes. Third, we were cognizant that NOWS outcomes could have differed whether the infant received clonidine or phenobarbital given that the former had to be weaned off prior to discharge. However, the use of any adjunctive represented a small number and did not alter our NOWS outcomes.

Conclusion

In conclusion, with the high prevalence of OUD in pregnancy, it is important for perinatal and neonatal providers to be cognizant of our findings to counsel families about the effects of different prenatal exposures on infant hospital NOWS outcomes. Existing literature typically reports hospital outcomes of prenatal opioid exposure alone and does not differentiate the types of opioid or exposure to additional neuromodulating substances. Our study encompasses multiple prenatal exposures in determining hospital NOWS outcomes. The knowledge gleaned from this study will provide anticipatory guidance and promote further research to understand inpatient outcomes for infants with opioid or polysubstance exposure.

Key Points.

• Hospital outcomes were similar between infants exposed to opioids alone or polysubstance including opioids.

• Infants with short- and long-acting opioids required pharmacologic treatment more often than either opioid alone.

• Differential hospital outcomes exist for various co-exposures of opioids with nonopioids.